Troposphere mapping function plays a very important role in space geodetic tropospheric correction, estimation, and application in meteorology. In the past decades, a series of mapping functions, such as NMF, GMF, VMF1 and VMF3, were successively developed for pursuing better mapping function performance. However, limited by modeling data source of numerical weather model, the mapping functions temporal-spatial resolutions are stuck in 6 h and 1°×1°, existing problems of bad applicability in extreme weather and regions with complex terrains. As the release of the generally acknowledged best global reanalysis ERA5, we have a chance to improve current mapping function modeling by taking full advantage of accuracy and temporal-spatial resolution (1h and 0.25°×0.25) of ERA5. In this contribution, we used the ERA5 reanalysis to establish the hourly site-wise Wuhan University Mapping Function (WMF) covering 1583 GNSS, VLBI and DORIS stations, and compared the accuracy and GNSS PPP performance of WMF with VMF3 at globally distributed stations. We found the significant accuracy improvements of WMF to VMF3 as well as the non-negligible vertical coordinate and ZTD difference biases between the two mapping functions.